Need help with depthai?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

luxonis
182 Stars 37 Forks MIT License 930 Commits 59 Opened issues

Description

DepthAI Python API utilities, examples, and tutorials.

Services available

!
?

Need anything else?

Contributors list

DepthAI Demo Program

This repo contains demo application, which can load different networks, create pipelines, record video, etc.

Documentation is available at https://docs.luxonis.com.

Python modules (Dependencies)

DepthAI Demo requires numpy, opencv-python and depthai. To get the versions of these packages you need for the program, use pip: (Make sure pip is upgraded:

python3 -m pip install -U pip
)
python3 install_requirements.py

Optional: For command line autocomplete when pressing TAB, only bash interpreter supported now: Add to .bashrc:

echo 'eval "$(register-python-argcomplete depthai_demo.py)"' >> ~/.bashrc

If you use any other interpreter: https://kislyuk.github.io/argcomplete/

Examples

python3 depthai_demo.py
- depth & CNN inference example

Conversion of existing trained models into Intel Movidius binary format

OpenVINO toolkit contains components which allow conversion of existing supported trained

Caffe
and
Tensorflow
models into Intel Movidius binary format through the Intermediate Representation (IR) format.

Example of the conversion: 1. First the

model_optimizer
tool will convert the model to IR format:
   cd /deployment_tools/model_optimizer
   python3 mo.py --model_name ResNet50 --output_dir ResNet50_IR_FP16 --framework tf --data_type FP16 --input_model inference_graph.pb

  • The command will produce the following files in the ResNet50_IR_FP16 directory:
    • ResNet50.bin - weights file;
    • ResNet50.xml - execution graph for the network;
    • ResNet50.mapping - mapping between layers in original public/custom model and layers within IR.
  1. The weights (

    .bin
    ) and graph (
    .xml
    ) files produced above (or from the Intel Model Zoo) will be required for building a blob file, with the help of the
    myriad_compile
    tool. When producing blobs, the following constraints must be applied:
  2. CMX-SLICES = 4 SHAVES = 4 INPUT-FORMATS = 8 OUTPUT-FORMATS = FP16/FP32 (host code for meta frame display should be updated accordingly)

    Example of command execution:

    /deploymenttools/inferenceengine/lib/intel64/myriadcompile -m ./ResNet50.xml -o ResNet50.blob -ip U8 -VPUMYRIADPLATFORM VPUMYRIAD2480 -VPUNUMBEROFSHAVES 4 -VPUNUMBEROFCMXSLICES 4

Reporting issues

We are actively developing the DepthAI framework, and it's crucial for us to know what kind of problems you are facing.
If you run into a problem, please follow the steps below and email [email protected]:

  1. Run
    log_system_information.sh
    and share the output from (
    log_system_information.txt
    ).
  2. Take a photo of a device you are using (or provide us a device model)
  3. Describe the expected results;
  4. Describe the actual running results (what you see after started your script with DepthAI)
  5. How you are using the DepthAI python API (code snippet, for example)
  6. Console output

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.